On Mon, 19 Oct 2015 01:52:55 -0500, "b2508" <108118@DSPRelated> wrote:>>On Sun, 18 Oct 2015 16:10:56 -0500, b2508 wrote: >> >>>>On Sun, 18 Oct 2015 12:07:04 -0500, b2508 wrote: >>>> >>>>> hi all, >>>>> is it possible to get real data by doing non complex downconversion? >>>>> assuming we receive real signal at some IF can you do conversion to >>>>> baseband by using just a sine or cosine? >>>> >>>>It depends on your signal design and what sort of information you need >>>>to recover from your signal. >>>> >>>>If you want all the information in a RF signal within some passband >>>>centered around that RF center frequency, you cannot mix down to "zero >>> IF" >>>>(that is, where the local oscillator is at the RF center frequency) >>>>without losing some information, unless you use a quadrature mixer, >>>>which gives you a complex output. >>>> >>>>If you need only certain information, such as the amplitude envelope, >>>>then you're okay without quadrature mixer, and with a real output at >>>>(so-called) baseband. This is a standard way of mixing down an >>>>AM-modulated signal. >>>> >>>>If you need all the signal information, but are okay with low-IF >instead >>>>of zero-IF, then you can also use a non-quadrature mixer. This is a >>>>standard way of mixing down an SSB-modulated signal. The result may >or >>>>may not be describable as "baseband" but as Tim points out that is >>>>terminology. The mixed-down signal is in a low band near DC. >>>> >>>> >>>>Steve >>> >>> How can I determine lowest IF to which I can move signal by just >>> multiplying it with sine or cosine? >>> >>> I have a real signal at 60 MHz originally and assuming that it has >>> bandwidth B and I want to preserve all of it, what would be the way to >>> calculate this lower IF? >> >>It's complicated, mostly by the fact that the edges of a "bandwidth" are > >>fuzzy, but also practical considerations in radio design. >> >>If 'B' is the bandwidth occupied by the signal plus sufficient guard >>bands such that images won't cause an issue, then the lowest "possible" >>IF is B/2, assuming you adequately filter the signal at 60MHz. >> >>Adequately filtering the signal with bandwidth B centered around 60MHz >>may be a challenge if it's not already in the digital domain and if B < >>6MHz -- this is why superheterodyne radios (and quadrature >downconversion) >>were invented. If you were going to do it "the old way" with LC filters > >>and single-LO (not quadrature) downconversion, you'd go down in IF by a >>factor of 10 each time until B >= 0.1 * carrier F, then you'd do one more > >>conversion. Or you'd design a crystal filter with bandwidth B (assuming > >>0 < B < 10kHz or so), then spend a large amount of money producing your >>filters. >> >>For that matter, adequately filtering the signal if it _is_ in the >>digital domain may be a challenge if B << 60MHz, because your filter >>length is roughly proportional to carrier/B (it's something like five >>times as long, probably, for a filter with halfway decent skirts, and >>longer if you need sharper skirts). So if B, for instance, is 600Hz then > >>you'd need a filter that's something like 50,000 taps long and that >>iterates at something well in excess of 60MHz. >> >>-- >>www.wescottdesign.com > >I am not sure I understood this. My signal is centered at 60 MHz in >digital domain and if it is wide, say 40 MHz, I also have an image of it >at the negative axes, right? What am I suppose to filter with this >filter?So it is already a real-valued digital signal with energy from about 40MHz to 80MHz? What is the sample rate?>From my point of view I could shift it to the left until some extent and >then filter out replicas or stuff that will not come into my signal (and >its replica on the negative side). When I try to move it to baseband I do >get this mixing of spectral components coming from positive and negative >side of frequency axes of original signal. >--------------------------------------- >Posted through http://www.DSPRelated.comAs others have pointed out, mixing with a real-valued sinusoid will produce both sum and difference frequencies. If the signal is already digitized this can produce significant problems with aliases, etc. Is there a reason you don't want to use a quadrature mix? That will strictly translate the entire spectrum by the frequency of the oscillator, without creating any new images or aliases. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
How to reconstruct original waveform from IQ
Started by ●November 12, 2008
Reply by ●October 19, 20152015-10-19
Reply by ●October 19, 20152015-10-19
On Mon, 19 Oct 2015 01:52:55 -0500, b2508 wrote:>>On Sun, 18 Oct 2015 16:10:56 -0500, b2508 wrote: >> >>>>On Sun, 18 Oct 2015 12:07:04 -0500, b2508 wrote: >>>> >>>>> hi all, >>>>> is it possible to get real data by doing non complex downconversion? >>>>> assuming we receive real signal at some IF can you do conversion to >>>>> baseband by using just a sine or cosine? >>>> >>>>It depends on your signal design and what sort of information you need >>>>to recover from your signal. >>>> >>>>If you want all the information in a RF signal within some passband >>>>centered around that RF center frequency, you cannot mix down to "zero >>> IF" >>>>(that is, where the local oscillator is at the RF center frequency) >>>>without losing some information, unless you use a quadrature mixer, >>>>which gives you a complex output. >>>> >>>>If you need only certain information, such as the amplitude envelope, >>>>then you're okay without quadrature mixer, and with a real output at >>>>(so-called) baseband. This is a standard way of mixing down an >>>>AM-modulated signal. >>>> >>>>If you need all the signal information, but are okay with low-IF > instead >>>>of zero-IF, then you can also use a non-quadrature mixer. This is a >>>>standard way of mixing down an SSB-modulated signal. The result may > or >>>>may not be describable as "baseband" but as Tim points out that is >>>>terminology. The mixed-down signal is in a low band near DC. >>>> >>>> >>>>Steve >>> >>> How can I determine lowest IF to which I can move signal by just >>> multiplying it with sine or cosine? >>> >>> I have a real signal at 60 MHz originally and assuming that it has >>> bandwidth B and I want to preserve all of it, what would be the way to >>> calculate this lower IF? >> >>It's complicated, mostly by the fact that the edges of a "bandwidth" are > >>fuzzy, but also practical considerations in radio design. >> >>If 'B' is the bandwidth occupied by the signal plus sufficient guard >>bands such that images won't cause an issue, then the lowest "possible" >>IF is B/2, assuming you adequately filter the signal at 60MHz. >> >>Adequately filtering the signal with bandwidth B centered around 60MHz >>may be a challenge if it's not already in the digital domain and if B < >>6MHz -- this is why superheterodyne radios (and quadrature > downconversion) >>were invented. If you were going to do it "the old way" with LC filters > >>and single-LO (not quadrature) downconversion, you'd go down in IF by a >>factor of 10 each time until B >= 0.1 * carrier F, then you'd do one >>more > >>conversion. Or you'd design a crystal filter with bandwidth B (assuming > >>0 < B < 10kHz or so), then spend a large amount of money producing your >>filters. >> >>For that matter, adequately filtering the signal if it _is_ in the >>digital domain may be a challenge if B << 60MHz, because your filter >>length is roughly proportional to carrier/B (it's something like five >>times as long, probably, for a filter with halfway decent skirts, and >>longer if you need sharper skirts). So if B, for instance, is 600Hz >>then > >>you'd need a filter that's something like 50,000 taps long and that >>iterates at something well in excess of 60MHz. >> >>-- >>www.wescottdesign.com > > I am not sure I understood this. My signal is centered at 60 MHz in > digital domain and if it is wide, say 40 MHz, I also have an image of it > at the negative axes, right? What am I suppose to filter with this > filter? > From my point of view I could shift it to the left until some extent and > then filter out replicas or stuff that will not come into my signal (and > its replica on the negative side). When I try to move it to baseband I > do get this mixing of spectral components coming from positive and > negative side of frequency axes of original signal.What are you really trying to do? You seem to be circling some unstated topic without ever getting right to it. Whether you downconvert to baseband or not is almost a tactical decision given the signal that you have -- why do you feel you need to, or why do you feel you don't want to? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●October 19, 20152015-10-19
Tim Wescott <seemywebsite@myfooter.really> wrote:>Whether you downconvert to baseband or not is almost a tactical decision >given the signal that you have -- why do you feel you need to, or why do >you feel you don't want to?Were Vlad here, he'd remind us that direct conversion to baseband is the "shit architecture" and you really want to use a superhet... S.
Reply by ●October 20, 20152015-10-20
On Mon, 19 Oct 2015 20:52:38 +0000, Steve Pope wrote:> Tim Wescott <seemywebsite@myfooter.really> wrote: > >>Whether you downconvert to baseband or not is almost a tactical decision >>given the signal that you have -- why do you feel you need to, or why do >>you feel you don't want to? > > Were Vlad here, he'd remind us that direct conversion to baseband is the > "shit architecture" and you really want to use a superhet...I'm not sure he would in this case -- if you're starting out in digital- land then all information is preserved. Trying to get a pair of mixers matched in phase and amplitude, not to mention getting the noise effects down, isn't a good game to play, and makes direct conversion problematical. -- www.wescottdesign.com
Reply by ●October 20, 20152015-10-20
I am trying to downconvert real signal that I get in digital domain centered at some IF frequency. Goal is to do it by not using quadrature downcconversion. From what I've seen from your explanations and some simulations, lowest IF to which I can move the signal is BW/2. After that I need to filter out components at the ends of spectrum... I was just trying to verify if this is correct and if filtering is done after multiplication with sine. --------------------------------------- Posted through http://www.DSPRelated.com
Reply by ●October 20, 20152015-10-20
Tim Wescott <tim@seemywebsite.com> wrote:> On Mon, 19 Oct 2015 20:52:38 +0000, Steve Pope wrote:>> Tim Wescott <seemywebsite@myfooter.really> wrote:>>>Whether you downconvert to baseband or not is almost a tactical decision >>>given the signal that you have -- why do you feel you need to, or why do >>>you feel you don't want to?>> Were Vlad here, he'd remind us that direct conversion to baseband is the >> "shit architecture" and you really want to use a superhet...> I'm not sure he would in this case -- if you're starting out in digital- > land then all information is preserved. Trying to get a pair of mixers > matched in phase and amplitude, not to mention getting the noise effects > down, isn't a good game to play, and makes direct conversion > problematical.But if you can get the original clock from the modulator to the demodulator, either pass it along, or extract it from the signal, then it should be pretty easy. Somewhat similar to using PLL to phase lock an analog demodulator. -- glen
Reply by ●October 21, 20152015-10-21
On Tuesday, October 20, 2015 at 4:30:30 AM UTC-4, glen herrmannsfeldt wrote:> Tim Wescott <tim@seemywebsite.com> wrote: > > On Mon, 19 Oct 2015 20:52:38 +0000, Steve Pope wrote: > > >> Tim Wescott <seemywebsite@myfooter.really> wrote: > > >>>Whether you downconvert to baseband or not is almost a tactical decision > >>>given the signal that you have -- why do you feel you need to, or why do > >>>you feel you don't want to? > > >> Were Vlad here, he'd remind us that direct conversion to baseband is the > >> "shit architecture" and you really want to use a superhet... >one implemetation difficulty with direct downconversion, is that the downconversion LO is on the same frequency as the desired channel. It can be difficult to implement hardware that is trying to recevice low level RF signal and also contains a high level source on the same frequency. Mark
Reply by ●October 21, 20152015-10-21
<makolber@yahoo.com> replies to my post,>> Were Vlad here, he'd remind us that direct conversion to baseband is the >> "shit architecture" and you really want to use a superhet...>one implemetation difficulty with direct downconversion, is that the >downconversion LO is on the same frequency as the desired channel. It >can be difficult to implement hardware that is trying to recevice low >level RF signal and also contains a high level source on the same >frequency.I agree. I usually state the problem more generally, that a mixer used for direct downconversion often must be (much) more mathematically accurate than a mixer used to mix to IF. Steve
Reply by ●October 21, 20152015-10-21
On Wed, 21 Oct 2015 12:40:47 -0700 (PDT), makolber@yahoo.com wrote:>On Tuesday, October 20, 2015 at 4:30:30 AM UTC-4, glen herrmannsfeldt wrote= >: >> Tim Wescott <tim@seemywebsite.com> wrote: >> > On Mon, 19 Oct 2015 20:52:38 +0000, Steve Pope wrote: >> =20 >> >> Tim Wescott <seemywebsite@myfooter.really> wrote: >> =20 >> >>>Whether you downconvert to baseband or not is almost a tactical decisi= >on >> >>>given the signal that you have -- why do you feel you need to, or why = >do >> >>>you feel you don't want to? >> =20 >> >> Were Vlad here, he'd remind us that direct conversion to baseband is t= >he >> >> "shit architecture" and you really want to use a superhet... >> =20 > >one implemetation difficulty with direct downconversion, is that the downco= >nversion LO is on the same frequency as the desired channel. It can be dif= >ficult to implement hardware that is trying to recevice low level RF signal= > and also contains a high level source on the same frequency. > >MarkAnd that usually appears as DC bleed-through, aka DC offset on the baseband output. This is why the center subcarriers of OFDM systems that use direct-conversion are not populated (e.g., 802.11). It is, however, often much cheaper to implement than a superhet architecture, and often provides a much broader practical tuning range. Sometimes the architecture has to fit the constraints of the requirements. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●October 21, 20152015-10-21
makolber@yahoo.com wrote: (snip on direct conversion)>> >> Were Vlad here, he'd remind us that direct conversion to baseband is the >> >> "shit architecture" and you really want to use a superhet...> one implemetation difficulty with direct downconversion, is that the > downconversion LO is on the same frequency as the desired channel. > It can be difficult to implement hardware that is trying to recevice low > level RF signal and also contains a high level source on the same frequency.Seems like it probably works better when you don't need to go all the way down to DC. Often enough what people call baseband doesn't go all the way, but close enough. -- glen
